Brain structural morphometry in non-musicians with superior pitch identification ability

Abstract

Absolute pitch (AP) has long served as a model for studying the neural correlates of pitch perception. Previous studies have examined brain function and structural morphometry in musicians with AP. However, pitch identification (PI) ability also exists in non-musicians, with some demonstrating superior PI ability. Despite this, little is known about their brain morphometry. The current study investigated structural differences between non-musicians with superior and average PI abilities by measuring gray matter volume and cortical thickness. Results revealed that, compared to non-musicians with average PI ability, non-musicians with superior PI ability had larger gray matter volumes–except in three regions where volumes were smaller–and smaller cortical thickness. These findings differ from prior studies of AP musicians, in which structural findings have varied that some reported smaller gray matter volumes and thinner cortex, while others report greater cortical thickness relative to non-AP musicians. The novel findings of the current study broaden our understanding of neuroanatomical correlates of PI ability, extending beyond AP musicians to include non-musicians with superior pitch identification skills.

Introduction

Pitch identification (PI) refers to the ability to recognize individual musical notes without an external reference (i.e., without first hearing or playing a known note)^1,2. This ability has been associated with certain musical tasks such as accompaniment, transcription, and instrument identification^3–5, although its practical utility in everyday musicianship is debated and may vary across musical genres and cultural contexts^6,7. Some studies have found that PI differs from other pitch perception abilities such as pitch contour perception, tone localization, and pitch discrimination, which are associated with distinct patterns of behavioral and neural activity^2,8.

The highest level of PI ability is absolute pitch (AP), which has been linked to cognitive advantages such as enhanced long-term pitch memory⁹, pitch-name recall even after interference¹⁰, semantic pitch matching¹¹, and tonal span memory¹². Neuroimaging studies have demonstrated several neural correlates of AP in musicians. For example, event-related potential (ERP) studies have shown that musicians with AP ability (AP musicians) exhibited reduced P300 amplitude, suggesting decreased reliance on working memory during PI tasks^13,14. Functional imaging studies have also provided insights into the neural correlates of AP. For instance, when comparing musicians with AP to those with relative pitch (RP) during tasks involving listening to musical tones or tone memorization, individuals with AP exhibit greater task-related activity in the left posterior dorsolateral frontal cortex, a region implicated in associative learning and memory^15–17. Furthermore, in contrasts designed to isolate pitch perception and labeling, elevated activity has been observed in the right planum temporale (PT), located in the posterior portion of the superior temporal gyrus, which plays a key role in pitch naming and auditory processing^17–19. Structurally, AP musicians have demonstrated greater left PT surface asymmetry associated with improved AP accuracy²⁰, possibly due to a smaller right PT rather than an enlarged left PT²¹.

However, structural studies of AP musicians have yielded mixed results. One study reported smaller gray matter volumes and thinner cortical thicknesses²², while another study found greater cortical thicknesses²³, in AP musicians. These discrepancies may be due to differences in sample size and/or the ratio of female to male participants.

Importantly, AP ability exists on a continuum from AP to quasi-AP (QAP), also referred to as weak- or pseudo-AP^19,22. QAP is typically characterized by slower and less accurate pitch processing compared to AP¹⁹. A positron emission tomography (PET) study of QAP musicians found activation of an extensive right hemispheric network, including the cerebellum (implicated in early auditory encoding, working memory, and multimodal encoding), the dorsolateral prefrontal cortex (associated with working memory), the middle and inferior frontal gyri (involved in motor/sensorimotor perception and semantics, respectively), and the superior and middle temporal gyri (associated with long-term semantic memory)¹⁹. These findings suggest that QAP musicians rely on multiple cognitive strategies and experience an increased pitch processing load relative to AP musicians. Although a left hemispheric advantage is still present, the right hemisphere may play a more prominent role in pitch processing for QAP musicians, a pattern that is not typically observed in individuals with true AP¹⁹.

Notably, PI ability is not exclusive to musicians. Several studies have shown that some non-musicians also demonstrate superior PI abilities^2,24,25, even though they may not be able to label or name notes without a reference. These individuals can reliably access or reproduce individual pitches. One behavioral study found that superior PI ability in non-musicians was associated with enhanced executive functions and mental rotation performance²⁶. Despite these findings, the neural correlates of PI ability in non-musicians have not been explored.

The current study aimed to examine brain structural morphometry in non-musicians with superior PI ability by measuring gray matter volume and cortical thickness. Gray matter volume was assessed using voxel-based morphometry (VBM). Classical VBM primarily yields maps of gray matter concentration or density, which can be confounded by volumetric changes introduced during spatial normalization spatial normalization²². To obtain a more direct measure of regional gray matter volume, we employed an optimized, modulated VBM pipeline implemented in the Computational Anatomy Toolbox (CAT12). This pipeline multiplies the spatially normalized gray matter segments by the non‑linear component of the Jacobian determinant derived from the registration to the MNI template. This modulation step corrects for local expansion or contraction caused by the normalization, thereby transforming the density maps into maps that approximate the absolute gray matter volume^27,28. Thus, the resulting “modulated, normalized, and smoothed” gray matter images used in our statistical analyses reflect regional gray matter volume rather than concentration. This modulated VBM approach is widely used in contemporary morphometric studies when the research question pertains to volume differences²⁹.

However, this method can risk fitting away meaningful signal variance²². In contrast, cortical thickness, which is derived from neuronal bodies, dendrites, glia and vasculature within the cortical layers, is more specific than VBM and is less susceptible to positional variance²³. However, it cannot capture subcortical gray matter structures²². Thus, the two measures are complementary to one another and together provide a more comprehensive understanding of brain morphometry.

We hypothesized that non-musicians with superior PI ability would exhibit distinct structural brain features—reflected in gray matter volume and cortical thickness—compared to non-musicians with average PI ability. Based on prior studies of musicians with absolute pitch, we expected differences in regions involved in auditory perception, memory, and sensorimotor integration. These include the middle temporal gyrus (linked to PI), orbitofrontal gyrus (involved in music memory and emotional evaluation), middle frontal gyrus (associated with motor and sensorimotor processing), and inferior parietal lobule (responsible for auditory memory, pitch judgement, visual-audio strategies).

Methods

Participants

Structural MRI and pitch identification data were collected from 320 undergraduates at Beijing Normal University (age range: 19–24 years; mean age = 20.45 ± 1.18 years; 191 females). All participants were right-handed, had normal or corrected-to-normal vision, and had no history of neurological or psychiatric disorders. The study was approved by the Institutional Review Board of the State Key Lab of Cognitive Neuroscience and Learning at Beijing Normal University and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants.

Of the initial sample, 34 participants (9 males, 25 females) demonstrated superior PI ability (PI accuracy > 80%²⁶). From the remaining 286 participants with average PI ability (accuracy rate < 80%), a control group of 34 individuals was selected, matched to the superior PI group on age and gender (9 males, 25 females). Only these 68 participants were included in the final analysis.

All participants were classified as “non‑musicians” based on the following criteria: (1) they were not music majors or enrolled in a formal conservatory; (2) they had no systematic, professional training in music theory or ear‑training; and (3) they were not professionally engaged in music. Despite this classification, some participants reported prior amateur instrumental experience (e.g., piano, violin). To account for this variability, musical training history was categorized by onset: early training (onset before age 6), late training (onset after age 9), or no training. Within the superior PI group, 20 had early training (mean starting age = 5.51 ± 0.66 years), 2 had late training (mean starting age = 12.70 ± 0.84 years), and 12 had no training. The matched average PI group contained 19 individuals with early training (mean starting age = 5.04 ± 0.75 years), 2 with late training (mean starting age = 12.85 ± 0.59 years), and 13 with no training. This categorical variable was included as a covariate in all statistical analyses. Detailed information on training duration and frequency was not collected, a limitation we acknowledge and discuss later.

Pitch identification test

The PI test was adapted from an absolute pitch test³⁰. Given the fact that the majority of participants had no formal musical training, only the seven basic notes of the C major scale in the fifth octave with piano timbre were used (i.e., C4, D4, E4, F4, G4, A4, and B4, with corresponding frequencies of 261.60, 293.66, 329.63, 349.23, 392.00, 440.00, and 494.88 Hz, respectively), played using a piano timbre. Participants completed a practice session with feedback, followed by a formal test without feedback. In the formal test, each of the seven pitches was presented three times in random order.

Each pitch was played twice for 500 ms, with an inter-stimulus interval of 1000 ms. Participants responded by clicking the corresponding pitch label on a computer screen. The test lasted approximately 10 min, including set-up and transition time. Both accuracy and reaction time were recorded, but only accuracy was analyzed in current study (see Table 1 for mean PI scores and standard deviations).

Table 1.

Characteristics and PI score of the participants.

Characteristics	Mean (SD)		t	p
	Superior PI	Average PI
Age (years)	20.23 (1.82)	20.55 (1.70)	−0.74	0.41
Gender (male/female)	34 (9/25)	34 (9/25)
Handedness	All right-handed
PI accuracy rate (%)	93.70 (6.81)	38.38 (5.96)	14.03****	0.0001

Standard deviations are shown in parentheses. ^****p <.001.

MRI data acquisition

Structural imaging was acquired using a 3 T Siemens MRI scanner at the Beijing Normal University MRI center. T1-weighted anatomical images were acquired with a 3D MPRAGE sequence: TR/TE/θ = 2530 ms/3.39 ms/7°, FOV = 256 × 256 mm², slice thickness/gap = 1.33/0 mm, in-plane resolution = 256 × 256, inversion time (TI) = 1100 ms, sagittal slices = 128. During the 6-min scan, participants laid supine on the scanner bed with foam pads to minimize head motion. Participants were instructed to close their eyes, remain still, and relax without focusing on anything in particular. Wakefulness during the scan was confirmed immediately after the scan via verbal inquiry, and no participants reported falling asleep during the scan.

Data preprocessing

T1-weighted data were preprocessed using the Computational Anatomy Toolbox (CAT12; http://www.neuro.uni-jena.de/cat/), which is a plug-in software based on Statistical Parametric Mapping (SPM12, https://www.fil.ion.ucl.ac.uk/spm/software/spm12/) and integrated into Matlab (Mathworks, Natick, MA, USA). CAT12 offers improved accuracy in gray matter volume analysis compared to previous VBM plug-in toolboxes based on SPM and is fully automated for surface-based analysis^31–34.

Preprocessing steps consisted of bias-field correction, skull-stripping, alignment to the Montreal Neurological Institute (MNI-152) structural template (classifying gray matter, white matter and cerebrospinal fluid) as well as spatial normalization with DARTEL registration (1.5 mm), all within the same generative model^29,34,35. Gray matter probability maps were then multiplied by the non-linear component of the Jacobian determinant, which corrects for volume changes introduced by spatial normalization and yields modulated images that represent local gray matter volume³⁶. Subsequently, a spherical harmonic method³⁷ was employed to reparametrize the cortical surface mesh based on an algorithm that reduces area distortions³⁸ to repair any topological defects^37–39. Finally, modulated gray matter probability maps were spatially smoothed with an 8 mm full-width at half maximum Gaussian kernel, and a spatially smoothed 15 mm FWHM Gaussian kernel was used when computing cortical thickness.

Statistical analysis

Group comparisons in gray matter volume and cortical thickness between those with superior and average PI abilities were conducted in CAT12. Musical training (early, late, and no) was considered as a covariate since the number of early or no musical training participants was relatively small (especially so for those with late musical training). Gender was also considered a covariate. Group differences were assessed using a non-parametric permutation approach (5,000 permutations) with threshold-free cluster enhancement (TFCE), applying a corrected threshold of p <.05.

Only regions exceeding 1,000 vertices for gray matter volume (as identified by the figure production toolbox xjView97, https://www.alivelearn.net/xjview/) and greater than 100 vertices for cortical thicknesses (cluster size × percentage covered in the specific region produced by CAT12) were considered statistically significant. Anatomical labels for gray matter volume were assigned using the Automated Anatomical Labelling (AAL) atlas, while the cortical thickness results were labeled using the Desikan–Killiany atlas (DK40)⁴⁰. Visualization was performed using xjView97 for volume and CAT12 for thickness. The values were not extracted from a single representative voxel. Instead, we extracted the mean value across the entire statistically significant cluster identified in our whole-brain group comparison. Specifically, for the gray matter volume, we calculated the mean modulated, normalized, and smoothed gray matter volume across all voxels within each significant cluster from the group comparison. For the cortical thickness, the mean thickness across all vertices within a significant cluster was automatically computed and exported by the CAT12 statistics module. This approach of using the mean across the entire cluster is the standard practice in voxel-based and surface-based morphometry, as it provides a more robust and representative measure of the structural characteristic of a region than a single peak voxel, while also increasing statistical power^41,42.

Additionally, for brain regions identified as significantly different between groups, the mean value across the entire significant cluster (gray matter volume or cortical thickness) was extracted. These values were then subjected to correlation analyses with PI score using Pearson’s correlation in SPSS (version 23). This approach of using the mean across a cluster is standard practice, as it provides a robust and representative measure of a region’s structural characteristic⁴¹.

Results

Consistent with the group assignment, the superior PI group demonstrated significantly higher pitch identification scores than the average PI group. Specifically, the superior PI group had a mean accuracy was 93.70 ± 6.81%, with scores ranging from 80.95% to 100%; notably, 15 participants achieved perfect scores. In contrast, the average PI group had a mean accuracy was 38.38 ± 5.96%, with scores ranging from 9.52% to 76.19% (see Table 1).

In terms of structural brain differences, the superior PI group showed larger gray matter volumes than the average PI group in several regions of the right hemisphere, including the middle temporal, middle orbitofrontal, and middle frontal gyri as well as the inferior parietal lobule. However, smaller gray matter volumes were also observed within the postcentral, superior occipital, and fusiform gyri in the same hemisphere. In the left hemisphere, the superior PI group exhibited larger gray matter volumes in the inferior orbitofrontal and middle temporal gyri as well as the cerebellum (see Table 2; Fig. 1).

Table 2.

Differences of gray matter volumes between non-musicians with superior and average PI ability.

Gray matter volumes	BA	Peak MNI coordinates			Cluster sizes in voxels	t	p
		x	y	Z
Right hemisphere
Middle temporal gyrus	20	33	15	−39	1942	7.999	< 0.001
Middle orbitofrontal gyrus	47	32	50	−14	2061	7.574	< 0.001
Inferior parietal lobule	39	48	−50	38	1022	5.719	0.003
Middle frontal gyrus	47	44	21	39	2011	9.662	< 0.001
Postcentral gyrus	3	45	−30	51	1845	−8.272	< 0.001
Superior occipital gyrus	19	25	−68	−2	1568	−7.483	< 0.001
Fusiform	39	50	−65	38	1683	−7.535	< 0.001
Left hemisphere							< 0.001
Cerebellum	37	−20	−45	−20	2708	9.310	< 0.001
Middle temporal gyrus	20	−60	−18	−23	1804	7.325	< 0.001
Inferior orbitofrontal gyrus	47	−47	32	−9	1350	6.511	< 0.001
Middle temporal gyrus	21	−51	−56	17	1152	5.846	0.003

The multiple comparison correction was used with non-parametric permutations (n = 5,000) and threshold-free cluster enhancement (TFCE) correction threshold p <.05 after 5,000 permutations. The results were labeled using the AAL atlas. Positive t value: superior PI larger than average PI. Negative t value: superior PI smaller than average PI. The p values are reported as p <.001 when below 0.001.

Fig. 1.

Differences of gray matter volumes between non-musicians with superior and average PI ability. The multiple comparison correction was used with non-parametric permutations (n = 5,000) and threshold-free cluster enhancement (TFCE) correction threshold p <.05 after 5,000 permutations. The results were produced by the xjView toolbox. Red and up arrow (↑): superior PI larger than average PI. Green and down arrow (↓): superior PI smaller than average PI. A: anterior; P: posterior; L: left; R: right. The results are visualized using the xjView97 toolbox (https://www.alivelearn.net/xjview).

Regarding cortical thickness, participants with superior PI ability showed consistently smaller cortical thickness compared to the average PI group. In the left hemisphere, this pattern was observed in the inferior temporal, lingual, and precuneus gyri. Across both hemispheres, the smaller cortical thickness was identified in the lateral occipital, superior temporal, middle temporal, and insula gyri, as well as the superior and inferior parietal lobules (see Table 3; Fig. 2).

Table 3.

Differences of cortical thicknesses between non-musicians with superior and average PI ability.

Cortical thicknesses	BA	Peak MNI coordinates			Cluster sizes in voxels	t	p
		x	y	z
Right hemisphere
Lateral occipital gyrus	18	29	−92	5	398	−4.345	0.021
Lateral occipital gyrus	18	15	−96	2	166	−3.927	0.028
Superior temporal gyrus	21	48	−39	−20	190	−6.448	0.010
Middle temporal gyrus	20	33	15	−39	126	−8.042	0.021
Superior parietal lobule	40	15	−47	72	246	−7.742	0.028
Inferior parietal lobule	7	18	−68	42	214	−7.711	0.028
Insula gyrus	47	30	20	−16	350	−6.387	0.010
Left hemisphere
Lateral occipital gyrus	18	−26	−87	−11	912	−7.391	0.009
Superior parietal lobule	40	−38	−44	58	618	−8.695	0.009
Inferior parietal lobule	7	−17	−71	41	547	−7.873	0.009
Superior temporal gyrus	21	−47	−30	−27	401	−6.346	0.010
Inferior temporal gyrus	20	−60	−18	−24	243	−7.109	0.009
Middle temporal gyrus	20	−60	−30	−8	213	−6.636	0.009
Lingual gyrus	37	−20	−45	−20	122	−8.905	0.009
Precuneus	5	−12	−50	69	152	−7.398	0.009
Insula gyrus	47	−28	20	−16	549	−6.554	0.010

The multiple comparison correction was used with non-parametric permutations (n = 5,000) and threshold-free cluster enhancement (TFCE) correction threshold p <.05 after 5,000 permutations. The results were labeled using the DK40 atlas. Superior PI non-musicians had all smaller cortical thicknesses than average PI non-musicians.

Fig. 2.

Differences of cortical thicknesses between non-musicians with superior and average PI ability. The multiple comparison correction was used with non-parametric permutations (n = 5,000) and threshold-free cluster enhancement (TFCE) correction threshold of p <.05 after 5,000 permutations. The results were produced by CAT12 toolbox. Blue: superior PI less than average PI. In the figure, the left hemisphere is in left side and the right hemisphere is in right side. The results are visualized using CAT12 toolbox (http://www.neuro.uni-jena.de/cat/).

In the correlation analyses of gray matter volume, the superior PI group showed significant negative correlations between PI performance and the right middle frontal gyrus and right middle temporal gyrus. In the average PI group, significant negative correlations were found between PI performance and the right middle orbitofrontal, right postcentral, left middle temporal, and left inferior orbitofrontal gyri. For cortical thickness, the superior PI group exhibited a significant positive correlation between PI performance and the left inferior temporal gyrus, while the average PI group showed a significant positive correlation between PI performance and the left insula gyrus (see Table 4).

Table 4.

Correlation analysis between PI performance and gray matter volume/cortical thickness in each group.

Measure	Group	Regions	r	p
Gray matter volume	Superior PI	Right middle frontal gyrus	−0.497	0.004
		Right middle temporal gyrus	−0.617	< 0.001
	Average PI	Right middle orbitofrontal gyrus	−0.398	0.024
		Right postcentral gyrus	−0.413	0.019
		Left middle temporal gyrus	−0.462	0.008
		Left inferior orbitofrontal gyrus	−0.447	0.010
Cortical thickness	Superior PI	Left inferior temporal gyrus	0.363	0.041
	Average PI	Left insula gyrus	0.357	0.045

The p values are reported as p <.001 when below 0.001.

Discussion

By comparing structural MRI measures, the current study uncovered neural correlates of superior PI in individuals lacking formal musical training. We found that individuals with superior PI ability exhibited a distinct pattern of brain morphometry compared to those with average PI ability: they showed significantly larger gray matter volumes in several brain regions—with the exception of three areas where volumes were smaller—and consistently smaller cortical thickness across multiple cortical regions. These volumetric and thickness differences were observed in several key areas, including the middle temporal gyrus, middle frontal gyrus, orbitofrontal gyrus, and inferior parietal lobule, aligning with our initial hypothesis.

Non-musicians with superior PI ability showed greater gray matter volumes in frontal regions, including the right middle frontal gyrus, the right middle orbitofrontal gyrus, and the left inferior orbitofrontal gyrus. Previous fMRI studies have implicated the orbitofrontal cortex in music memory⁴³, auditory sensitivity⁴⁴, spatial perception⁴⁵, and the evaluation and processing of emotions evoked by musical stimuli^46–51. The middle frontal gyrus has been associated with motor and sensorimotor perception⁵². In prior studies, AP individuals have been found to employ a variety of cognitive strategies—including verbal encoding⁵³, semantic associative memory, internal “emotional” representations¹¹, and auditory, visual, kinesthetic, and spatial imagery^10,19—during AP identification. For example, some AP individuals reported retaining pitch information using verbal labels⁵³ or verbal encoding strategies¹⁰. Others described using emotional labels—such as “unnatural”, “poignant”, “achy”, or “bitter” to identify the pitch “F^#”, illustrating the use of semantic or emotional associations. Still others use spatial position or visual imagery¹¹. Similarly, Wilson et al.¹⁹, found that some QAP musicians utilized strategies such as timbre perception, pitch location on the keyboard, or visualizing finger positions to identify pitches. These findings suggest that individuals with superior PI—like AP musicians—may similarly engage multiple processing strategies. Future research should investigate how specific orbitofrontal subregions contribute to high-level PI performance.

In addition to frontal regions, the superior PI group exhibited significantly larger gray matter volumes in the bilateral middle temporal gyri, right inferior parietal lobule, and left cerebellum. These findings are consistent with prior work indicating that AP musicians possess a prominent leftward middle temporal gyrus, which is important for pitch identification^19,54. The inferior parietal lobule has been implicated in auditory memory⁵⁵ and pitch judgement tasks⁵⁶, which may involve visual-mental strategies⁵⁷. While traditionally associated with motor function, the cerebellum is increasingly recognized for its roles in auditory tasks and pitch memory^55,58,59 as well as higher-order cognitive functions such as multimodal encoding and working memory^60–62. The cerebellum may thus contribute to both pitch discrimination and memory in individuals with superior PI ability^1,55.

Interestingly, some regions showed smaller gray matter volumes in the superior PI group, including the left superior occipital, fusiform, and postcentral gyri. Both the superior occipital and fusiform gyri, while primarily visual regions, have been associated with visual-auditory processing relevant to music perception, although no visual input is involved during PI performance^17,63. The postcentral gyrus is involved in sensorimotor perception and multisensory integration⁵⁴. Although the reason for these smaller gray matter volumes remain unclear, one structural MRI study found a similar pattern in occipital and postcentral regions in trained musicians, which was interpreted as reflecting greater neural efficiency and automaticity due to filtering of irrelevant sensory inputs during musical performance⁶⁴. While this interpretation may not fully account for our findings, a similar process of neural refinement or functional specialization could be at play in non-musicians with superior PI ability, who may have developed highly effective pitch-processing strategies. It should be noted, however, that such interpretations are speculative, and alternative explanations—including developmental differences or reduced synaptic density—cannot be ruled out.

The observed thinner cortex in the superior PI group may likewise reflect similar adaptive or efficiency-related mechanisms, though the direction of structural differences does not inherently imply superior function without corroborating functional evidence. Notably, the superior temporal gyrus—a key region in pitch discrimination, pitch memory, and auditory frequency processing⁶⁵—showed thinner cortex in the superior PI group. This region is also linked to phonological and semantic processing^22,66, and AP musicians have been found to exhibit leftward asymmetry in the superior temporal gyrus and associated white matter^20,22,23,54. The inferior temporal gyrus, involved in music-related visual imagery, is similarly affected⁶⁷.

The involvement of the right superior and middle temporal gyri aligns with prior findings in quasi-absolute pitch (QAP), who are thought to rely on reference tones to make interval judgements¹⁹. These same strategies may support superior PI in non-musicians.

Within the parietal lobe, the superior parietal lobule is involved in spatial cognition, mental rotation, and music pitch-to-space transformations⁶⁸. The inferior parietal lobule, as noted earlier, supports auditory memory and pitch discrimination. The insula also appears to be critical, as it contributes to sensorimotor control, learned motor representations^69,70, and potentially verbal labeling during PI⁷¹. Thinner cortex of the insula has been observed in AP musicians⁷², further suggesting its importance. The precuneus, part of the default mode network, is implicated in music-related imagery, memory, emotion, and visual perception^73,74, while the lingual gyrus is involved in visual encoding of tonal information⁷⁵. Altogether, these findings support the idea that both AP and superior PI rely on a diverse set of cognitive strategies and distributed neural networks.

Our results revealed a pattern in which superior PI non-musicians had larger gray matter volumes in many regions (except for the postcentral, superior occipital and fusiform gyri in the right hemisphere) but consistently thinner cortex compared to average PI individuals. This pattern is somewhat inconsistent with previous studies: Bermudez et al.²², reported smaller gray matter volumes and thinner cortex in AP musicians, while Dohn et al.²³, observed greater cortical thicknesses. These inconsistencies may arise from methodological differences such as sample size (71 in Bermudez et al., 34 in Dohn et al.,), test materials (self-developed in-house AP test in Bermudez et al., pitch identification test (PIT) in Dohn et al., or gender distribution (47 females in Bermudez et al., 28 males in Dohn et al.). In the current study, the total group (superior and average PI) consisted of 18 males and 50 females, which resembles the gender distribution in Bermudez et al., and may help explain the observed pattern of smaller cortical thickness.

Our findings should also be considered in light of recent large-scale, multi-modal neuroimaging studies of pitch expertise, particularly the work of Lutz Jäncke and colleagues at the University of Zurich^76–79. Their research program, which has extensively examined structural, functional, and connectivity differences between musicians with AP and RP, provides a comprehensive framework for understanding specialized auditory abilities. A key insight from this body of work is the “neural efficiency” hypothesis, where long-term, specialized training can lead to more streamlined neural processing⁷⁶. For example, a foundational diffusion tensor imaging study demonstrated that AP musicians possess reduced fractional anisotropy in the arcuate fasciculus compared to RP musicians, interpreted as a sign of more efficient, specialized fronto-temporal communication for automatic pitch labeling⁷⁷. This aligns with the widespread cortical thinning we observed in superior PI non-musicians, potentially indicating a similar trend towards efficient processing, even in the absence of formal musical training. More recent work further elucidates this by showing how AP is associated with distinct patterns of anatomical and functional connectivity^78,79. Conversely, our finding of increased gray matter volume in specific frontal and temporal regions may reflect a different kind of neuroanatomical adaptation—one that supports the diverse, multi-strategy cognitive approaches (e.g., visuo-spatial, semantic, or sensorimotor associations) we hypothesize are employed by non-musicians without formal pitch-labeling training. The Zurich group’s work underscores that superior pitch processing involves a complex reconfiguration of brain structure and networks. Our study extends this principle to the non-musician individuals, suggesting that the neural signatures of high pitch ability may manifest differently depending on the developmental path (explicit training vs. informal strategy use).

Our correlation analyses showed that PI performance was significantly associated with both gray matter volume and cortical thickness in the superior and average PI groups. Among individuals with superior PI, performance was negatively correlated with gray matter volume in the right middle frontal gyrus and in the right middle temporal gyrus, regions involved in motor/sensorimotor perception and pitch identification, respectively. Additionally, a positive correlation was observed between PI performance and cortical thickness in the left inferior temporal gyrus, a region associated with music-related visual imagery. In contrast, individuals in the average PI group exhibited negative correlations between PI performance and gray matter volume in the orbitofrontal gyrus (linked to music memory and emotion), postcentral gyrus (sensorimotor perception and multisensory integration), and the left middle temporal gyrus. A positive correlation was also observed between PI performance and cortical thickness in the insular cortex, which is implicated in sensorimotor control, learned motor representations, and potentially verbal labeling during PI tasks. Although some brain regions with significant group differences had no significant correlations with PI performance, they may still contribute to the cognitive mechanisms underlying PI in each group.

An intriguing pattern emerged from the conjunction of the group comparison and correlation analyses. While the superior PI group exhibited significantly larger gray matter volumes in the right middle frontal gyrus and right middle temporal gyrus compared to the average PI group, within the superior PI group, higher PI accuracy was negatively correlated with gray matter volumes in these same regions. This apparent dissociation suggests a complex, potentially non-linear relationship between brain structure and behavioral proficiency. One plausible interpretation is that larger gray matter volumes in these frontal and temporal regions may represent a neuroanatomical trait that facilitates the development of superior PI ability, perhaps supporting the cognitive strategies—such as sensorimotor integration, working memory, or associative learning—discussed earlier. However, among individuals who have already attained a high level of ability, those with the very highest performance scores may undergo a process of progressive neural refinement or efficiency. In this framework, a relative reduction in gray matter volume among the top performers could reflect more streamlined neural circuits, reduced synaptic pruning, or greater automaticity in pitch processing, akin to the neural efficiency hypotheses observed in other cognitive domains^64,80. Thus, the overall group difference may reflect a structural foundation that enables superior PI ability, whereas the within-group correlation suggests a distinct, performance-dependent process of neural refinement. This dissociation highlights that group differences and within-group behavioral gradients can reflect separate neurostructural relationships, both of which are necessary for a comprehensive understanding of brain-behavior associations in specialized abilities.

It is important to note that the superior PI group in the current study included individuals with early, late, and no musical training. Prior studies have shown that AP is more likely to develop in individuals with early musical training, especially if they also have a genetic disposition, while late musical training is less likely to result in actual AP ability^{19,30,81–83}. However, early training alone is not sufficient; a gene × environment interaction likely underlies AP acquisition (Hou et al., 2016). AP also shows familial aggregation, further supporting a genetic basis^82–84. Some studies have even gone as far as to link AP ability to specific genetic markers on chromosome 2 and 6q^85,86. However, while early musical training can facilitate AP development, it is not necessary—many individuals, including those in the current study, demonstrate superior PI ability without it^25,87. Further research is needed to determine whether the neural mechanisms underlying PI differ between musicians and non-musicians and whether the same genetic factors apply.

Some limitations of this study should be noted, however. First, some participants had early (before the age of 6) or late (after the age of 9) experience with amateur instrumental learning. However, as the data were collected many years ago, detailed information regarding training duration, frequency, or cumulative hours was not systematically recorded. To account for this, we included musical training experience (categorized as early, late, or no training) as a covariate in all statistical models and matched the distribution of training onset types between groups. Second, and critically, our measure of musical background remains coarse. The lack of detailed, quantitative data on cumulative musical exposure—such as total training hours, frequency of practice, or engagement in informal music activities—is a significant constraint. To decisively separate the neuroanatomical correlates of an innate “ability trait” from those shaped by “training effects”, future studies should employ more precise assessments of musical experience using standardized instruments like the Goldsmiths Music Sophistication Index (Gold-MSI). Systematically collecting such granular data and analyzing its correlation with both pitch identification performance and brain structure will be essential. Only with this level of precision can we clarify whether the structural differences observed here predominantly reflect a predisposed neural architecture for superior pitch processing, are a consequence of informal but cumulatively significant musical engagement, or result from an interaction between the two. Third, the PI test in present study only used the seven basic pitches of C major and was answered by clicking on the screen tags, and participants were not required to have knowledge of pitch names or a background in music theory. The practice and feedback session before the formal test is designed to ensure the understanding of the task. Therefore, we believe that the test scores mainly reflect the PI ability itself rather than the results of musical training. Despite this, when interpreting the results, we should remain cautious and acknowledge that the heterogeneity of musical exposure among the participant groups is an important variable that requires in-depth exploration in future research. Fourth, our results are correlational, and thus we cannot infer causality between PI ability and gray matter volumes or cortical thicknesses. Fifth, it is important to consider that musical experience, even informal, likely exists on a continuum. Our operational definition of ‘non-musicians’—based on the absence of formal, professional training—allowed us to study pitch identification ability outside the context of musical expertise. However, this classification may encompass a range of informal exposures (e.g., casual instrument playing, singing, attentive listening) that were not fully quantified. Such a continuum of informal experience could contribute to the development or maintenance of superior PI performance. Therefore, our findings should be interpreted within the specific context of our participant criteria. They illuminate the neuroanatomical correlates associated with superior PI ability in individuals who lack formal musical training, but they may not fully disentangle the effects of latent ability from those of varied, informal musical engagement. Future studies employing more granular assessments of lifelong musical exposure will be crucial to clarify these relationships. Sixth, our sample was restricted to native Chinese participants who spoke a highly tonal language. Prior research has shown that tonal language speakers (those who speak Mandarin, Cantonese, Thai, Vietnamese) are more likely to possess AP-like abilities, potentially due to the pitch sensitivity required in their everyday speech^88–90. While tonal language speakers may not use the same pitch labeling strategies as musicians, they demonstrate enhanced long-term pitch memory^9,25,91–93. Future studies should explore whether the neural correlates of PI differ between tonal and non-tonal language speakers, and between musicians and non-musicians.

Conclusion

The current study examined structural brain differences between non-musicians with average and superior pitch identification ability, specifically focusing on gray matter volume and cortical thickness. Compared to those with average PI ability, individuals with superior PI exhibited significantly larger gray matter volumes in several brain regions (with the exception of three areas showing smaller volumes) and consistently smaller cortical thickness. These findings suggest that superior PI ability in non-musicians may be supported by a combination of structural features that facilitate multiple cognitive strategies.

A key insight emerges from the conjunction of group comparisons and within-group correlations. While the superior PI group showed larger gray matter volumes in frontal and temporal regions, higher performance within this group was associated with reduced volume in these same areas. This dissociation points to a complex, non-linear relationship between brain structure and behavior: a certain structural foundation may enable the development of superior ability, but the highest levels of proficiency may be marked by neural refinement or efficiency. This finding underscores the importance of moving beyond simple group comparisons to examine how brain-behavior relationships vary across the spectrum of ability.

Overall, the current findings provide novel insights into the neural basis of pitch perception, extending our understanding beyond trained musicians to include non-musicians with superior PI ability. They highlight that specialized perceptual skills can emerge outside formal training and are associated with distinct, and sometimes counterintuitive, patterns of brain structure.

Author contributions

J.H. designed and performed the experiment, analyzed the data, wrote and revised the main manuscript. J.H. and T.H. wrote the revised the manuscript. M. O. supervised and revised the mansucript. All authors reviewed the manuscript.

Funding

This study was supported by the Fujian Normal University Research Start-Up Funding (Y0720304K05), and the 111 Project from the Ministry of Education of China (B07008).

Data availability

The data that support this study are available from the corresponding author, Jiancheng Hou, upon reasonable request. Contact email: bonjovi_hou@163.com.

Declarations

Competing interests

The authors declare no competing interests.